You trained a binary classifier model which gives high accuracy on a very large test dataset, but a much lower accuracy on the training data. This outcome could reasonably be explained by:

$($Select all that apply.$)$

$$ The training may not have been well regularized.

$$You may have used the test examples during training or validation.

$$This could be an instance of underfitting.

$$The training and test data could have been sampled from different distributions.

$$This could be an instance of overfitting.